some modbus slaves are lossing connection periodically

[DCIM_Support]

I'm monitoring a Circutor modbus devices. There are 27 slaves with the same modbus IP address and periodically some slaves are lossing connection.

(CID:116361027)

[DCIM_Support]

Hi Enric Climent,

Unfortunately, at such number of Modbus slave devices it is normal 😀. I often meet similar problems. There are my recommendations:

1. reduce number of Modbus slave devices on one IP address (to 5-8);
2. reduce the frequency poll of Modbus slave devices (one poll in 2-3 minutes);
3. increase the number of attempts of poll for each of problem Modbus slave devices (not 3, and 5-6 attempts – for DCE-7.3.1 and newer);
4. increase timeout between unsuccessful attempts poll of Modbus slave devices (for example, up to 15 seconds – for DCE-7.3.1 and newer).

As option it is possible to reduce number of sensors, which are read out on Modbus slave devices.

And at last, to try to understand, and probably to reveal and eliminate defects of the Modbus serial communication line between Modbus devices. Practically, the last recommendation can be the most effective and useful.

With best regards.

(CID:116361050)

[DCIM_Support · Accepted solution]

Hi Enric Climent,

Yesterday officially released the new DCE-7.4.3: StruxureWare Data Center Expert v7.4.3 Release Notes. And as it is written in the release:

Improvements to Modbus polling
Modbus polling is no longer affected by prolonged network disruptions.

Therefore, this may be another recommendation in the form of item number 5. Try it.

With best regards.

P.S.: In addition, here more detailed information from Jackie Lehr on this news subject on my question:

Q: Explain, please, in more detail, what do you mean? How is this achieved?

A: There was an issue where Modbus polling consumed resources if network connectivity was interrupted for an extended period. It is now fixed.

(CID:116920091)

[DCIM_Support]

This question is closed for comments. You're welcome to start a new topic if you have further comments on this issue.